The discovery of microRNAs (miRNAs) and other short non-coding RNAs (such as siRNA, piRNA, and snRNA) has led to a rapid expansion of research elucidating their expression and diverse biological functions. These functions include regulation of development, cell proliferation, differentiation, and the cell cycle as well as translation or stability of target mRNAs (Zamore & Haley 2005; Bushati & Cohen 2007). The DNA sequences encoding miRNAs are transcribed by RNA polymerase II as long pri-miRNAs that are processed, first by Drosha into pre-miRNA (70-90 nt) and then by Dicer to yield mature miRNAs consisting of 21-22 nt single strands (Bartel 2004; Zamore & Haley 2005; Kim & Nam 2006). Hundreds of miRNAs have been identified, which may cooperatively regulate greater than one-third of all human genes (Lewis et al. 2005; Lim et al. 2005; Kim & Nam 2006). Recent studies have shown that distinct miRNA expression patterns are associated with various types of cancer (Lu et al. 2005a; Cummins & Velculescu 2006; Esquela-Kerscher & Slack 2006; Hammond 2006a; Pfeffer & Voinnet 2006; Hernando et al. 2007; Wu et al. 2007) and viral infections (Cullen 2006; Dykxhoorn 2007; Pan et al. 2007). Thus, miRNAs may be considered as potential diagnostic biomarkers as well as potential drug targets for antisense agents (Hutvagner et al. 2004; Meister et al. 2004; Krutzfeldt et al. 2005; Davis et al. 2006; Hammond 2006b; Orom et al. 2006; Weiler et al. 2006; Zhang & Farwell 2008).
In most cases, expression levels of many different miRNA species (rather than a single miRNA) are changed in the course of disease, and therefore should be assayed simultaneously for monitoring progression of the disease and response to therapy. For example 27 human miRNAs are either down-regulated (let-7a, let-7b, let-7c, let-7d, let-7g, miR-16, miR-23a, miR-23b, miR-26a, miR-92, miR-99a, miR-103, miR-125a, miR-125b, miR-143, miR-145, miR-195, miR-16, mir199a, miR-221, miR-222, miR-497) or up-regulated (miR-202, miR-210, miR-296, miR-320, miR-370, mir498, miR-503) in prostate carcinoma (Mattie et al. 2006; Porkka et al. 2007). Cancer biopsies are often formalin fixed, which is incompatible with mRNA isolation and analysis due to the RNA-protein cross-links, covalent modifications and degradation of long RNA that occur during the fixation process. However, short miRNAs have significant advantages as biomarkers because they are much less affected by such modifications (Li et al. 2007; Xi et al. 2007).
It is commonly expected that many more naturally occurring small RNAs are still to be discovered. Once identified and validated as biomarkers and/or therapeutic targets, specific miRNA(s) require sensitive and accurate detection and quantification in biological and clinical samples. The copy number of individual miRNAs may vary from less than 10 to about 50,000 per cell, and their expression profiles vary with the age, health and treatment of cells and where they are in the cell cycle (Chen et al. 2005; Ahmed 2007). A variety of methods for measuring the levels of known miRNAs have been already developed, including but not limited to: Northern blots (Valoczi et al. 2004; Aravin & Tuschl 2005; Ramkissoon et al. 2006; Pall et al. 2007); nuclease-protection (Lee et al. 2002; Overhoff et al. 2004; Aravin & Tuschl 2005; Winkler et al. 2006); DNA primer-extension (Seitz et al. 2004; Sioud & Rosok 2004); sandwich hybridization assays using ELISA and DNA dendrimers (Barad et al. 2004; Lu et al. 2005a; Mora & Getta 2006); direct labeling of miRNAs and hybridization to slide or bead arrays (Krichevsky et al. 2003; Babak et al. 2004; Barad et al. 2004; Calin et al. 2004; Liu et al. 2004; Nelson et al. 2004; Shingara et al. 2005; Yeung et al. 2005; Xia 2006; Gottwein et al. 2007; Tang et al. 2007; Wang et al. 2007); pre-amplification and labeling of target sequences and hybridization to slide or bead arrays (Saba & Booth 2006; Mattie et al. 2006); RT-PCR with TaqMan detection probes (Chen et al. 2005; Jiang et al. 2005; Jacobsen et al. 2005; Lu et al. 2005b; Raymond et al. 2005; Winkler et al. 2006); ligation-assisted PCR wherein miRNA serves as a ligation splint (Brandis et al. 2006; Sorge & Mullinax 2006; Maroney et al. 2007; Chamnongpol & Souret 2008), the Invader assay (Allawi et al. 2004); rolling circle amplification (RCA) of target-specific padlock probes (Jonstrup et al. 2006; Van Huffel 2006); and single miRNA molecule detection based on hybridization with short LNA-DNA probes (Neely et al. 2006).
Many of these methods have been adapted from previously established mRNA assays with modifications that accommodate the differences between mRNA and miRNA. MicroRNAs are much smaller than mRNAs and are neither capped nor polyadenylated. These characteristics make it hard to isolate a pure fraction of miRNAs, limit the number of labels that can be chemically or enzymatically introduced into miRNA molecules, and disallow the use of standard PCR primers (see below).
Northern hybridization of miRNA targets with labeled oligonucleotide probes is still considered as the “gold standard” for the simultaneous characterization of miRNAs and their longer precursors (pri- and pre-miRNAs). This method, however, is inadequate for a number of reasons. First, short unmodified RNAs cannot be efficiently cross-linked to support membranes used in nucleic acid hybridization assays. Second, even when cross linking occurs there is significant variability dependent on the number of U residues present; and third, cross-linked species have reduced ability to hybridize with probes (Valoczi et al. 2004; Pall et al. 2007). Two other methods that rely on gel-electrophoresis techniques are: nuclease-protection of labeled DNA or probes, which are usually longer than target miRNA targets; and reverse-transcription extension of primers, which are usually shorter than miRNA targets. The major limitations of all three of these methods are poor sensitivity, preventing detection of low-copy miRNAs, and low throughput and multiplexing capabilities.
Some other methods of miRNA detection employ direct chemical or enzymatic modification of the RNAs (Wark et al. 2008). For example, platinum (Babak et al. 2004) and alkylating (Enos et al. 2007) agents that preferentially bind purine bases (G>A) are used for chemical labeling of miRNAs. Drawbacks of these approaches include: (1) efficacy of labeling depends on number and position of the purine bases, which vary for different miRNA species; (2) the number of introduced labels into the same miRNA species may vary; and (3) these modifications may reduce the affinity of the miRNA for probes (Ahmed 2007). Alternatively, modification of miRNAs can be made through oxidation of the 2′,3′-diol on their 3′ termini (Liang et al. 2005; Beuvink et al. 2007), but it is a laborious, multi-step procedure and also causes partial degradation of the RNA.
Enzymatic approaches applied to miRNA modification involve either RNA or DNA ligases. miRNA labeling methods usually involve derivatives of pCp and T4 RNA ligase (Cameron & Uhlenbeck 1977) and their efficiency varies depending on last few nucleotides located at the 3′ end of the miRNA (Cao 2004; Esquela-Kerscher & Slack 2004; Nelson et al. 2004; Enos et al. 2007; Wang et al. 2007). Moreover, T4 RNA ligase tends to circularize small RNAs naturally carrying 5′-p and 3′-OH, including miRNAs (Aravin & Tuschl 2005; Nichols et al. 2008). The RNA circularization prevents 3′-end labeling or adapter attachment (another approach involving T4 RNA ligase), and, therefore, makes the majority of small RNA undetectable by current ligation-based methods. To prevent circularization, the 5′-p ends of RNAs must be dephosphorylated. The application of T4 RNA ligase for attachment of adapter or linkers to small RNA does not require prior knowledge of the RNA sequence and, therefore, is used mostly for discovery of new small RNAs. For this purpose, universal linkers are attached at each end of small RNAs followed by conventional PCR amplification, cloning, and sequencing (Aravin & Tuschl 2005; Pfeffer et al. 2005; Cummins et al. 2006; Michael 2006). A similar approach was also applied for expression profiling of miRNAs, wherein asymmetric PCR was used after conventional PCR and the resulting single-stranded PCR products were hybridized to target-specific oligonucleotide probes attached to color-coded beads (Lu et al. 2005a).
Splint-assisted ligation by T4 DNA ligase is an alternative strategy for the attachment of adapter oligonucleotides to target RNA. This reaction is not very efficient for ligation of RNA (rather than DNA) between the 3′-OH and 5′-p ends and its efficiency depends on the sequence near the ends to be ligated (Moore & Query 2000). Hence the use of ligases to label miRNAs may lead to biased representations of the different miRNAs species. Similar to T4 DNA ligase, T4 RNA ligase 2 ligates RNA and DNA ends but in both splint-assisted and splint-independent manners (Ho & Shuman 2004; Nandakumar & Shuman 2004; Nichols et al. 2008). Another commonly used enzymatic technique is polyadenylation of miRNAs at their 3′ ends (Aravin & Tuschl 2005; Shi & Chiang 2005; Ahmed 2007; Wark et al. 2008), which allows the use a DNA polymerase primer extension with oligo(dT)-primers. This technique is of limited value, however, because polyadenylation-based assays cannot detect all miRNA species (Enos et al. 2007). Both sequence and structure of the RNA may affect the poly(A) polymerase processivity (Yehudai-Resheff et al. 2000). Moreover, poly(A) polymerase cannot extend RNAs having a 2′-OMe modification on their 3′ terminal nucleotides (Ebhardt et al. 2005; Yang et al. 2006) and, therefore, polyadenylation cannot be currently used for detection of such RNAs (Enos et al. 2007). Although the 2′-OMe modification is not typical for animal miRNAs, this modification is common for plant siRNAs and miRNAs as well as for piRNAs from Drosophila and animals (Li et al. 2005; Yu et al. 2005; Aravin et al. 2007; Norwich et al. 2007; Yang et al. 2007).
RNAs are known to serve as primers for nucleic acid polymerization. A variety of DNA polymerases, reverse transcriptases and mutated RNA polymerases can catalyze the polymerization of DNA using RNA primers with both DNA and RNA templates. For example, the Klenow fragment of DNA polymerase I has been used for the selective labeling and detection of specific RNAs in a mixture (Huang & Szostak 1996, 2003; Huang & Alsaidi 2003). RNA-primed array-based Klenow enzyme assays (RAKE) have been used for labeling (during primer extension) and detection of miRNAs by hybridization to DNA capture probes attached either to microarray slides (Nelson et al. 2004; Yeung et al. 2005; Berezikov et al. 2006; Getts et al. 2006; Genisphere 2007) or beads (Jacobsen et al. 2005). Also, the ability of miRNAs to serve as primers was employed for miRNA detection in vitro using circular DNA probes and RCA (Jonstrup et al. 2006; Van Huffel 2006) as well as for detection of miRNA in situ using ultramer DNA probes (Nuovo et al. 2009). Finally, RNA-dependent extension of the miRNA by both DNA (reverse transcriptase) and RNA (RdRp) polymerases has been used for identification of mRNA sequences targeted by these miRNAs by (Rana 2004; Vatolin et al. 2006).
Target RNA can serve as a template for RT-PCR. The major problem for direct RT-PCR of small RNAs is related to their size—at only 20-27 nt, they are nearly the same size as an ordinary PCR primer while two primers are required for exponential amplification. For this reason, the first RT-PCR assays were developed for miRNA precursors, which are more than twice as long as mature miRNAs (Schmittgen et al. 2004; Jiang et al. 2005). This methodology, however, may not lead to accurate representation of the biologically relevant profile, because levels of the precursors do not always correlate with those of the mature miRNAs due to the rapid processing of the miRNA precursors and the longevity of active miRNAs associated with the RISC complex (Bartel 2004; Jiang et al. 2005; Lao et al. 2007).
One approach for amplifying miRNA relies on having only a short overlap between primer and miRNA sequences (Chen et al. 2005; Raymond et al. 2005; Raymond 2007; Sharbati-Tehrani et al. 2008; Sharbati-Tehrani & Einspanier 2008). Unfortunately, ordinary short RT-PCR primers fail to hybridize stably at the temperatures needed for the PCR extension step. Moreover, miRNA sequences differ significantly in GC-content, both among different miRNA species (in the range 24 to 73% GC) and between the 5′ and 3′ halves of individual miRNAs (Hammond 2006c). As a result, primers binding to different miRNAs would not be equally effective under given conditions, compromising both sequence-specificity and efficacy of the PCR amplification (Esquela-Kerscher & Slack 2004; Winkler et al. 2006). An additional problem is the difficulty of distinguishing minor differences in sequence and/or length between different miRNA isoforms (Sioud & Rosok 2004; Hammond 2006c). To address these problems, extended target-specific primers forming short but stable duplexes with the 3′ ends of miRNAs have been used by three groups. One group used stem-and-loop RT primers having only 6 nt complementary to the 3′ end of target miRNAs along with two more PCR primers, wherein one primer was corresponding to the 5′-end of target sequence and another primer was corresponding to the stem-and-loop sequence, and TaqMan probes (Chen et al. 2005). The second group used combo RT primers, which comprised 7-12 nt complementary to the 3′ end of target miRNAs and an additional sequence encoding sequence for second PCR primer, along with two more PCR primers, wherein one LNA-DNA primer corresponded to the 5′-end of target sequence and the other primer corresponded to the additional sequence (Raymond et al. 2005; Raymond 2007). And third group used similar combo RT primers (but without LNA modifications) along with three additional PCR primers, wherein one PCR primer corresponding to the 5′-end of target sequence was the combo primer and other two primers corresponded to the additional sequences of the combo primers (Sharbati-Tehrani et al. 2008; Sharbati-Tehrani & Einspanier 2008).
A second approach for assaying short mature miRNA by RT-PCR is the extension of short target sequences (miRNA or complementary cDNAs) either by polyadenylation (Shi & Chiang 2005; Illumina 2007; Kreutz et al. 2007), or ligation of adapter oligonucleotides (Lu et al. 2005b; Dawson & Womble 2006; Mishima et al. 2007).
Simultaneous amplification of many target sequences in one reaction under the same conditions (multiplex PCR) could increase assay throughput and allow the use of smaller samples. However, reported multiplex PCR reactions are restricted to amplification of five to ten targets (Broude et al., 2001). The reasons for this are that conventional PCR primers specific to different targets tend to form dead-end dimers when mixed and extended together, and there is also increased cross-hybridization of primers with non-target sequences (Brownie et al. 1997). As a result, primer design for multiplex PCR is not a trivial task, requiring tedious optimization of PCR conditions and it still often fails—especially for short RNA targets with high variation of GC-contents such as miRNAs. There is an added technological challenge because of overlap in the emission spectra of available fluorescent dyes. Currently at most six dyes can be assayed simultaneously within the same sample. One approach to achieve uniform multiplex PCR amplification is using combo primers, wherein each combo primer combines different pairs of target-specific sequence and an additional Zip-code sequence. PCR with combo primers is usually performed in two rounds of amplification: the first round is performed with a relatively low concentration of the combo primers while the second round uses a high concentration of shorter primers comprising only the Zip-code sequences. A wide variety of 20-27 nt Zip-code or functionally similar sequences can be associated with (designated to) targets of interest (Gerry et al. 1999; Ye et al. 2001; Fan et al. 2000; Hirschhorn et al. 2000). Such sequences share several common features: (1) they are designed to be unique, not represented in the sample to be tested; (2) have similar Tm so that annealing and extension can be performed under the same stringent condition; and (3) do not cross-hybridize to each other or to another or nucleic acids that can be present in a sample (Shoemaker et al. 2006; Smith et al. 2001; Shuber et al. 2005; Lin et al. 2006; Pinto et al. 2006).
RNA size and sequence play key roles in any RNA detection method relying on a sequence-specific binding (hybridization) of target RNAs either with substantially complementary capture probes or primers. The differences in thermostability between perfect and mismatched duplexes depend on length and sequence as well as the type and position of mismatches. The trade-off between high affinity for the target and low sequence-specificity of binding is a major limitation for designing allele-specific hybridization probes and multiplex probes targeting sequences with different GC-content (Toulme et al. 2001; Demidov & Frank-Kamenetskii 2004). Increasing the affinity of these agents to their intended targets simultaneously decreases their selectivity. Hybridization and primer-extension assays dealing with individual sequences can be optimized for maximum selectivity by adjusting temperature, incubation time, salt, and formamide concentration in the hybridization and washing steps. However, multiplexing assays, in which multiple probe-target hybridizations are conducted simultaneously under the same conditions, lack this customizing option. There are so-called stringency elements known in art that can improve sequence-specificity of hybridization probes and primers including: (1) modified nucleotides (e.g. LNA), which provide higher affinity to AT-rich sequences, placed into specific positions in the probe/primer sequence (Braasch et al. 2002; Valoczi et al. 2004; Fluiter et al. 2005); (2) dividing of probes/primers into smaller fragments that are complementary to adjacent sites in target RNA (Maher & Dolnick 1988; Kandimalla et al. 1995; Wang et al. 2003); (3) stem-loop (hairpin) structures with short single-stranded overhangs complementary to the target 3′-end, which enhance stability through contiguous stacking interactions between the probe and target ends (Lane et al. 1998; Chen et al., 2005; Wang et al. 2007); (4) partially double-stranded probe/primers that bind with target in through competitive or replacement hybridization process (Vary 1987; Li et al. 2002; Kong et al. 2004; Huang et al. 2007; Luk et al. 2007); (5) folding-back sequences that are complementary to one or to both ends of the probe/primer sequence (Roberts & Crothers 1991; Hertel et al. 1998; Ohmichi & Kool 2000; Bortolin & Zastawny 2007); (6) “molecular beacon”-like structures that have short complementary “arms” flanking the antisense sequence at both ends (Bonnet et al. 1999, Hartig et al. 2004; Hopkins & Woodson 2005); and (7) use of substantially complementary sequences that have few mismatches to the intended target in specific positions (Guo et al. 1997; Delihas et al. 1997; Brukner et al. 2007).
The rapidly expanding list of different proprietary methods of miRNA detection indicates that no current technology is perfect or has clear advantage over its competitors. Because RT-PCR methods have very good sensitivity, sequence specificity, and dynamic range, they are frequently used as method of choice for expression profiling of defined miRNAs as well as validating results obtained by other common methods such as microarray and northern blot assays (Ahmed 2007). However, none of these methods is particularly simple, with most requiring numerous steps that render them laborious, time consuming and expensive. Moreover, the need for cumbersome and costly temperature cycling equipment limits the wide adoption of even PCR-based methods for point-of-care diagnostic applications, an area in which isothermal amplification techniques could provide simpler and more cost-effective solutions.
Before the present invention, the circularization of small RNAs naturally carrying 5′-p and 3′-OH by T4 RNA ligase was regarded as an obstacle and explicitly avoided (Aravin & Tuschl 2005) while short lengths of small RNA targets was recognized as factor limiting the use of conventional PCR primers in current assays (Chen et al. 2005; Jiang et al. 2005). Aspects of the present invention include methods and compositions for detection of known small RNAs as well as the discovery of new small RNAs. We capitalize on the ability of small RNA targets or their conjugates with oligonucleotide adapters to be easily circularized. The circular RNA templates provide amplification of the target (and adapter) sequences via synthesis of multimer nucleic acids that can be either labeled for direct detection or subjected to PCR amplification and detection. The structure of small circular RNAs and their corresponding multimer nucleic acids provide certain advantages, including unmatched flexibility in design of conventional RT and PCR primers as well as allowing the use of overlapping dimer-primers for efficient and sequence-specific amplification of short target sequences. As compared to previously described methods, aspects of the present invention allow a reduction in the number of steps and reagents while increasing sensitivity and accuracy of detection of small RNAs with both 2′OH and 2′-OMe at their 3′ ends.